Emmanuel Ogu

Key Skills

Software

Labelbox

Label Studio

Mindrift

SuperAnnotate

Toloka

CVAT

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Computer Code Programming

Video

Top Task Types

Data Collection

Evaluation/Rating

RLHF

Freelancer Overview

I have hands-on experience in AI training data and annotation, specializing in evaluating model responses, validating JSON/function-calling workflows, and ensuring high-quality instruction-following behavior. My work includes applying detailed rubrics such as truthfulness, completeness, harmfulness, formatting, and user-intent understanding to assess LLM outputs. I’ve also contributed to projects involving multi-step reasoning evaluation, edge-case detection, and dataset quality checks for generative AI systems. Beyond annotation, I bring a strong QA mindset including functional testing, scenario validation, and attention to detail which helps me identify inconsistencies, ambiguities, and data-quality gaps quickly. I’m comfortable working with structured guidelines, maintaining consistency across large datasets, and delivering high-accuracy labeling under time constraints. My combined experience in QA, API testing, and AI evaluation allows me to approach data labeling with both analytical rigor and user-focused perspective.

Labeling Experience

AI trainer

CVATImageObject DetectionEvaluation Rating

The 4 Evaluation Axis (commonly used for AI image or completion comparison tasks.) Instruction Following (IF) → How well the output follows the prompt or instructions. ✅ Check: Did the image or text reflect what was explicitly asked? ❌ Penalize: Missing or ignored instructions, partial edits, or adding irrelevant elements. Image Consistency (IC) → How well it maintains consistency with the original image or reference. ✅ Check: Identity, layout, style, color palette, or subject details remain intact. ❌ Penalize: Unnecessary changes, character drift, mismatched angles, or off-model details. Quality (Q) → The overall visual or technical quality of the result. ✅ Check: Good rendering, clear details, proper lighting, and realistic textures. ❌ Penalize: Artifacts, distortions, messy edges, or unnatural compositions. AI-ness / Naturalness (AN) → How natural or realistic the image looks; freedom from “AI artifac

The 4 Evaluation Axis (commonly used for AI image or completion comparison tasks.) Instruction Following (IF) → How well the output follows the prompt or instructions. ✅ Check: Did the image or text reflect what was explicitly asked? ❌ Penalize: Missing or ignored instructions, partial edits, or adding irrelevant elements. Image Consistency (IC) → How well it maintains consistency with the original image or reference. ✅ Check: Identity, layout, style, color palette, or subject details remain intact. ❌ Penalize: Unnecessary changes, character drift, mismatched angles, or off-model details. Quality (Q) → The overall visual or technical quality of the result. ✅ Check: Good rendering, clear details, proper lighting, and realistic textures. ❌ Penalize: Artifacts, distortions, messy edges, or unnatural compositions. AI-ness / Naturalness (AN) → How natural or realistic the image looks; freedom from “AI artifac

2025

Education

Work History